This is the start of breaking the main dump109.c file into smaller
modules to make it a bit more maintainable.
Move all the #define and structure declarations into dump1090.h
Changes based on ideas from John VK1ET. His commit notes are as follows
:
1. Change input socket handling to avoid unnecessary memmove (use
pointers instead)
2. Add ability to read and decode binary beast format TCP.
3. Change output socket handling same as sbs_output - only call output
handler if there is a current client connected use separate ports for
beast in and out - no need for --net-beast flag. Will input/output
format defined by socket connected to. avr raw, beast binary and sbs can
be handled simultaneously.
4. Some comments changes, filtering of ModeAC to json,.
Resolves some merge conflicts that I couldn't find a way to fix during
the previous merge.
Remove a few Tabs that escaped in the edit/merge
Rename the fix_errors flag variable to nFix_crc since it now represents
the number of crc errors to attempt to fix.
Remove the aggressive variable since it's now part of nFix_crc
Only print phase correction statistics when phase correction is enabled.
Add an additional Modes.stat_ph_bit_fix[] to count the bits fixed during
phase correction
Sorry Ulrich - I can't get Github to resolve the merge errors and
preserve your commit notes, so I'll add them here.
Improvements on bit error correction, doc update, preparation for
program installation/package build
Hi,
I committed some further improvements on the bit error correction code,
updated the readme, and implemented a way to install the program in the
linux file system hierarchy (allows for package building).
Regards,
Ulrich
Bug fixes in the SBS output code as a result of an EMAIL from Liviu
Some software gets upset if null fields are output where the software is
expecting a valid value. This was being caused by DF0's outputting and
SBS MSG5. MSG5 needs Atert and SPI flags, which are derived from the FS
field. DF0 doesn't have an FS field, so cannot output a MSG5. Change so
that DF0 results in an SBS MSG7.
Also DF17/DF18 metypes 5 to 18 contain raw Lat/Lon data. However they
cannot be decoded into real Lat/Lon values unless both even and odd
parts are available Therefore, when we receive a DF17/DF18 we need to
check that we have successfully decoded the Lat/Lon before we attempt to
send an SBS MSG2 or MSG3. If we don't have a decoded Lat/Lon, send a
MSG7 instead.
Phase enhancement is used to try and increase the signal amplitude when
Nyquist sample aliasing is suspected. In previous versions, this was
enabled by default. In this version, the default is off. There is now an
additional command line switch --phase-enhance to turn it on.
There are also additional debug statistics which count the number of
phase demodulation failures in for both uncorrected and phase corrected
passes.
Ideally we should clear down the mm structure every loop of our
ModeA/C/S bit detector. However, we're getting 2Mbits of data per
second, and the structure is several tens of bytes long. Clearing down
every loop would require us to zero up to 100Mb per second. The memset
function may be fast, but it's still going to take up valuable processor
time.
So instead of clearing the whole structure every loop, just clear the
important parts.
Instead of force clearing mm.crc to zero if we successfully correct bit
errors, just use the mm.correctedbits variable. This allows us to print
out the crc value containing the errors during list output modes.
Actually use the (DF17) messages where we do bit correction.
Other changes to AVR message I/O.
1) treat mlat timestamp of zero as invalid (easier to work with than
(-1) and equally unlikely).
2) If we don't have a valid timestamp send a without timestamp
'*.......' message rather than '@.......' with invalid time.
3) Drop interpretation of obsolete AVR ' #' & '$' formats - they
wouldn't have worked anyway (wrong length).
If we error correct a DF17 frame, check that the ICAOaddr exists in our
recently seen ICAO cache. This reduces the likelyhood of birthday
paradox solutions producing false ICAO addresses
Restrict Syndromes to exclude DF bits. Do NOT introduce syndromes for
errors in the first 5 bits (DF field)
Use the --aggressive flag to see if we populate the two-bit error
syndromes. If --aggressive is not specified, then we don't attempt to
fix two bit errors.
vk1etI believe there is an error in the way the CRC error syndromes are
being created in the two bit case. The inner loop was introducing an
extra error bit each time through rather than moving it. Below is the
modified code fragment for syndrome creation collapsedSun 07:26
19 May 2013 07:26vk1et [notifications@github.com]Actions
To:Mantirez/dump1090 [dump1090@noreply.github.com]Cc:MMalcolmRobb
[Support@ATTAvionics.com]
To help protect your privacy, some content in this message has been
blocked. If you're sure this message is from a trusted sender and you
want to re-enable the blocked features, click here.
Bugfix first noted by VK1ET, and posted here :
https://github.com/antirez/dump1090/pull/23#issuecomment-18113094
I believe there is an error in the way the CRC error syndromes are being
created in the two bit case. The inner loop was introducing an extra
error bit each time through rather than moving it.
Updated the version number for Ulrich's code
Replaced some Tabs with spaces
Forward declaration of the fixBitErrors() function required for M$ VC
6.0 to suppress warning messages.
Based on a submission by VK1ET, but I've tweaked it a bit to use the
bFlags.
Basically, only print out values in --interactive that have been
received from the aircraft. Previous versions would display 0.000 for
lat and long for any positionless aircraft. This version prints blanks
instead for any/all fields that are not yet known.
Remove dependency of the SBS output code on the historic (a) aircraft
structure. The only items that were required were the decoded aircraft
Lat/Lon and these are now included in the mm structure.
Rewrite the SBS output code to use mm->bFlags when populating the output
fields. This ensures that all available data is output, and also that no
stale data is sent.
Using the mm->bFlags variable for SBS output means there is no further
requirement for the sbsFlags member in the aircraft structure, so remove
it.
Cross your fingers and hope this hasn't introduced too many bugs !
Implement additional bFlags to validate AircraftOnGround, and FS.
Decode Aircraft On Ground from DF0, DF4, DF5, DF16, DF20, and DF21 where
possible.
Flag FS as valid for DF4, DF5, DF20, DF21,
Remove dr and um from the mm structure. They're only used in console
list output mode, so decode them there if required.
Additional flags in the bFlags variables for even and odd CPR lat/lon
updates.
Change the interactiveReceiveData function to use the bFlags when
updating the aircraft (a) structure from the newly received (mm) message
structure. This should hopefully be faster than basically re-decoding
the DF, type and subtypes all over again.
If we decode the lat/lon into the aircraft (a) structure, back populate
it into the message (mm) structure. This allows us to print a decoded
Lat/Lon in the decoded message list output.
If --modeac is specified, then the program will be building an aircraft
list. Calculate the altitude when we create the aircraft structure (a) ,
and then back copy it into the mm structure. This avoids us having to do
a second ModeAtoModeC conversion in the list output display.
Also - bug fix in the ns_vel calculation.
Changes to 'dump1090.c'-file made accordingly.
modified: .gitignore
modified: dump1090.c
deleted: gmap.html
new file: public_html/gmap.html
new file: public_html/script.js
new file: public_html/style.css
Update the bFlags structure member to indicate which other structure
members contain decoded values.
Also, trim out mm structure members that are simple bitwise ands from
the raw data, unless the results are used in lots of places whereby
decoding them once is more efficient.
Extend the DF-17 CPR decoding to use User Lat/Lon for ground positions.
Also start implementing a bFlags structure variable to indicate which
fields in the mm and a data structures contain valid values.
Based on initial code supplied by VK1ET
Allow the user to specify their receivers (technically their aerials)
physical location. This is required for decoding ground positions, and
also for uploading aircraft data to various sharer sites.
The position can be hardwired by setting the MODES_USER_LATITUDE_DFLT
and MODES_USER_LONGITUDE_DFLT constants before compilation, or entered
at runtime via the new command line switches --lat and --lon. Either
way, the values are checked for validity before use.
Only clear mm after we are sure we have got a Mode A/C. This reduces
processor load.
Also tidy up some Linux compiler warnings that result from the last
change.
Based on a submission by VK1ET. Fully decode DF-17, metype 19. Take care
to only update fields that are valid in the data.
Also, initialise (to zero) the mm structures before use
1) Implement additional timestamps to remember when a new Lat/Lon value
is generated.
2) Increase callsign[] character array to 16 chars to make it long word
aligned.
3) Change some more comments from /* to //
Use memset to clear entire structures at initial creation to zero, and
then populate values that should not be zero. Do this rather than
individually clear each structure member to zero.
This should be faster, plus it's safer when new structure members are
added because you don't have to remember to initialise them.
